26.01.27 - A computational Perspective on the challenge of inner interpretrability - Fede Adolfi

Mech Interp → Understanding interpretable processes inside a complex object (i.e. nn). Should be a useful process, an example is circuit discovery